OpenAdapt.AI
FREE TO RUN LOCALLY · MIT OPEN SOURCE1.7k stars on OpenAdapt· 259 forks

Automate the UI-only work your APIs can’t reach.

OpenAdapt compiles demonstrations into one governed loop across browser, desktop, RDP, and Citrix. Browser runs in production today; desktop, RDP, and Citrix run through customer-controlled qualification. It verifies consequential results and halts when it cannot prove the intended outcome.

Verified business effects · fail-closed execution · customer-controlled sensitive data

pip install openadaptMIT licensedNo account or Cloud required

OpenAdaptObservingStep 1 of 33Observing the application
0:00 / 0:00Evidence

Guided view synchronized to the exact retained runtime timeline; raw footage remains unchanged.

  1. 1
    DemonstrateCapture the task and its evidence.
  2. 2
    ExecuteReplay the compiled workflow locally.
  3. 3
    Verify or haltProve the effect, or stop for review.
Three real applicationsOpenEMR · Frappe Lending · openIMIS
One execution contractDemonstrate → replay → verify or halt
Healthy compiled runs0 model calls

Customer result

Recovering missed billables with automated RVU audits

OpenAdapt automated the repetitive EMR navigation and programmatic reconciliation behind a cardiology electrophysiology practice’s monthly RVU audits, then produced a review workbook and a ready-to-send recovery email, with each read-back verified before it counted.

A US cardiology electrophysiology practice
Monthly physician RVU recovery audit
Partner engagement; institutional endorsement pending

Read the case study

≈$75,000

estimated recoverable billables identified per year

≈5 hrs

manual audit work saved each month

0

silent incorrect writes reported as done

How it works

From demonstration to verified operation

One lifecycle connects qualification, execution, evidence, and governed change.

  1. 1

    Demonstrate and qualify

    Show one bounded task. Review its program, risks, identities, effects, fault cases, and supported environment before certifying it.

  2. 2

    Deploy in your boundary

    Run approved browser workloads in managed execution or keep sensitive desktop, RDP, Citrix, private-network, and restricted-egress work customer-controlled.

  3. 3

    Execute, verify, and repair

    Healthy runs use no generative-model calls. OpenAdapt proves the declared effect, halts on uncertainty, and subjects every candidate repair to the workflow contract.

The right fit

Built for consequential UI-only work

Use the simplest reliable tool. OpenAdapt fits when the interface is unavoidable and the outcome must be proved.

Use OpenAdapt when

  • Inputs are mostly structured and the business intent stays stable.
  • The target application has no practical write API for the last mile.
  • A wrong action has operational, financial, or compliance cost.
  • The resulting state can be checked independently or by reacquiring the persisted record.
  • The workflow repeats often enough to justify qualification.

Do not use OpenAdapt when

  • A reliable supported API already completes the work.
  • The task depends on open-ended judgment or changes constantly.
  • No meaningful success condition can be defined.
  • The task is low-volume, low-consequence, and ordinary scripting already works.

See the complete decision guide for APIs, browser automation, RPA, computer-use agents, scripts, and OpenAdapt on the comparison page.

Cross-surface execution

One governance model across the interfaces you use

Portable workflow intent stays separate from each environment-specific binding, so DOM selectors, accessibility elements, and remote visual anchors are never treated as interchangeable.

Browser runs in production today; desktop, RDP, and Citrix run through customer-controlled qualification. Each surface carries its own measured acceptance evidence, published per surface in the qualification evidence.

Browser

Combine DOM, accessibility, visual, and interaction evidence for governed web execution.

Native desktop

Operate Windows, macOS, and Linux applications through native accessibility plus retained visual evidence.

Remote applications

Drive RDP, Citrix Workspace, and VDI externally through pixels, keyboard, and mouse while preserving identity, policy, verification, and audit.

Choose the execution boundary

Managed execution

Use the OpenAdapt control plane and managed runners for approved browser and non-sensitive workloads.

See Cloud pricing

Customer-controlled deployment

Run sensitive data, native applications, RDP, Citrix, private networks, and restricted-egress workflows inside your boundary. External black-box remote execution does not require software inside the managed session.

Review the security boundary

Commercial offer

Qualify the workflow before scaling the deployment

OpenAdapt Community

Free and MIT licensed. Record, compile, qualify, and run locally with no account or Cloud service.

Workflow Qualification Sprint

From $15,000

In a ten-business-day target, we assess one named workflow, application, environment, identity contract, effect verifier, failure boundary, deployment path, and ROI. You receive a qualified prototype and a signed go/no-go report.

Complex native, RDP, and Citrix scopes are typically $25,000–$40,000.

Qualify one workflow

OpenAdapt Cloud

$500.00/month

Up to 10,000 workflow runs/month

A separate managed subscription for approved workflows, run history, evidence, and governed updates. Enterprise qualification, regulated deployment, and SLAs are scoped separately.

See Cloud and enterprise pricing

OpenAdapt Cloud

From approved workflow to reviewable outcome.

This is the hosted product running today at app.openadapt.ai. Workflows, runs, evidence, and reports stay connected in one reviewable dashboard.

The OpenAdapt Cloud workflows dashboard at app.openadapt.ai: approved workflows, compiled bundle versions, and run history in the real hosted product.

Dashboard. Approved workflows, compiled bundle versions, and run history in one hosted dashboard.

Real OpenAdapt Cloud interface

Trust by construction

Keep control of the data, action, and evidence

  • Customer-controlled execution for sensitive and remote work
  • Local processing for live screenshots, OCR, identities, and secrets
  • Explicit network and model policy for every production profile
  • Signed, sealed, versioned, and independently verifiable artifacts
  • Evidence-backed VERIFIED or HALTED outcomes
  • MIT-licensed local runtime with a commercial enterprise control plane

Open the trust center

Strengthen the commons

Early access

Contribute data for credits

Every sanitized contribution strengthens the shared hardening corpus, the commons that lowers everyone's silent-wrong-effect rate. In return, you earn run credits that extend your usage allowance.

  • Sanitized, de-identified derivatives only. Raw recordings never leave your machine.
  • You approve every byte through hash-bound local review before anything is shared.
  • Opt-in and off by default. You can stop future contributions at any time.
  • Every contribution must meet the named de-identification standard in the versioned terms, and your organization attests before sharing.
Request access to the contributor programEarly access. Opt-in. Sanitized derivatives only.

Bring the real workflow

Qualify one workflow

Bring the actual application, environment, workflow, and success condition. We will determine whether it should be automated and what verified production requires.

Start the qualification